Since I have worked with Oracle I have removed duplicates using the rowid column. SQL Server has something similar but its an undocumented feature (might be changed without notice).

delete

from table t1

where exists

(select *

from table t2

where t1.column1 = t2.column1

and t1.column2 = t2.column2

...

and t1.rowid > t2. rowid) -- in Oracle

and t1.%%physloc%% > %%physloc%%) -- in SQL Server

This self join can be useful, you dont need to compare all rows for example and it removed more than one row in case of tripple duplicates.